Spaces:
Sleeping
Sleeping
import pandas as pd | |
import numpy as np | |
from sklearn.model_selection import train_test_split, GridSearchCV | |
from sklearn.linear_model import SGDClassifier | |
from sklearn.metrics import classification_report, confusion_matrix, make_scorer, f1_score | |
import shap | |
import xgboost as xgb | |
import gradio as gr | |
import matplotlib.pyplot as plt | |
import joblib | |
SVM = joblib.load('SVM.pkl') | |
Log_Reg = joblib.load('Log_Reg.pkl') | |
XGB = xgb.XGBClassifier() | |
XGB.load_model('XGB.model') | |
df = pd.read_csv('Superstore.csv') | |
df.dropna(subset=["Region", "Category", "Sub-Category", "Quantity", "Discount"], inplace=True) | |
MEDIAN = 8.662 # from the exploratory analysis file | |
RANDOM_STATE = 42 # random seed to ensure results are reproducible | |
region=np.unique(df['Region'], return_inverse=True)[1] | |
category=np.unique(df['Category'], return_inverse=True)[1] | |
subCategory=np.unique(df['Sub-Category'], return_inverse=True)[1] | |
# turn quantity, discount, and profit columns into vectors of numbers | |
quantity = df["Quantity"].to_numpy() | |
discount = df["Discount"].to_numpy() | |
profit = df["Profit"].to_numpy() | |
vectorizedDataset = np.empty((len(region), 5)) | |
labels = np.empty(len(region)) | |
# generate feature vectors | |
for i in range(0, len(region)): | |
data = np.zeros((1, 5)) | |
data[0][0] = region[i] | |
data[0][1] = category[i] | |
data[0][2] = subCategory[i] | |
data[0][3] = quantity[i] | |
data[0][4] = discount[i] | |
vectorizedDataset[i] = data | |
if (profit[i] > MEDIAN): | |
labels[i] = 1 | |
else: | |
labels[i] = 0 | |
train, test, trainLabels, testLabels = train_test_split(vectorizedDataset, labels, test_size=0.3, random_state=RANDOM_STATE) | |
region_label = {'Central': 0, 'East': 1, 'South': 2, 'West': 3} | |
category_label = {'Furniture': 0, 'Office Supplies': 1, 'Technology': 2} | |
sub_category_label = {'Accessories': 0, 'Appliances': 1, 'Art': 2, 'Binders': 3, 'Bookcases': 4, | |
'Chairs': 5, 'Copiers': 6, 'Envelopes': 7, 'Fasteners': 8, 'Furnishings': 9, | |
'Labels': 10, 'Machines': 11, 'Paper': 12, 'Phones': 13, 'Storage': 14, 'Supplies': 15, | |
'Tables': 16} | |
profit_label = {0: 'Below Median Profit', 1: 'Above Median Profit'} | |
feature_names = ["Region", "Category", "Sub-Category", "Quantity", "Discount"] | |
def sanitize_inputs(Region, Category, Sub_Category, Quantity, Discount): | |
try: | |
Region = region_label[Region] | |
Category = category_label[Category] | |
Sub_Category = sub_category_label[Sub_Category] | |
except KeyError: | |
return ["Please provide region, category, and sub category from the pre-defined Superstore dataset classes", None] | |
if Quantity < 1 or Discount < 0: | |
return ["Quantity and Discount must be positive", None] | |
if not isinstance(Quantity, int): | |
return ["Quantity must be an integer", None] | |
if Discount > 1: | |
return ["Discount cannot be greater than one", None] | |
return [Region, Category, Sub_Category] | |
def XGB_predict(Region, Category, Sub_Category, Quantity, Discount): | |
sanitized = sanitize_inputs(Region, Category, Sub_Category, Quantity, Discount) | |
if len(sanitized)==2: | |
return sanitized | |
input = np.array([[sanitized[0], sanitized[1], sanitized[2], Quantity, Discount]]) | |
predicted_class = XGB.predict(input) | |
explainer = shap.Explainer(XGB, test) | |
shap_values = explainer(input) | |
shap_values.feature_names = ["Region", "Category", "Sub-Category", "Quantity", "Discount"] | |
plot = shap.plots.bar(shap_values, show=False) | |
plt.savefig('shap_plot_XGB.png') | |
return [profit_label[predicted_class[0]], 'shap_plot_XGB.png'] | |
def SVM_predict(Region, Category, Sub_Category, Quantity, Discount): | |
sanitized = sanitize_inputs(Region, Category, Sub_Category, Quantity, Discount) | |
if len(sanitized)==2: | |
return sanitized | |
input = np.array([[sanitized[0], sanitized[1], sanitized[2], Quantity, Discount]]) | |
predicted_class = SVM.predict(input) | |
explainer = shap.Explainer(SVM, test) | |
shap_values = explainer(input) | |
shap_values.feature_names = ["Region", "Category", "Sub-Category", "Quantity", "Discount"] | |
plot = shap.plots.bar(shap_values, show=False) | |
plt.savefig('shap_plot_SVM.png') | |
return [profit_label[predicted_class[0]], 'shap_plot_SVM.png'] | |
def Log_reg_predict(Region, Category, Sub_Category, Quantity, Discount): | |
sanitized = sanitize_inputs(Region, Category, Sub_Category, Quantity, Discount) | |
if len(sanitized)==2: | |
return sanitized | |
input = np.array([[sanitized[0], sanitized[1], sanitized[2], Quantity, Discount]]) | |
predicted_class = Log_Reg.predict(input) | |
explainer = shap.Explainer(Log_Reg, test) | |
shap_values = explainer(input) | |
shap_values.feature_names = ["Region", "Category", "Sub-Category", "Quantity", "Discount"] | |
plot = shap.plots.bar(shap_values, show=False) | |
plt.savefig('shap_plot_LogReg.png') | |
return [profit_label[predicted_class[0]], 'shap_plot_LogReg.png'] | |
LogReg_tab = gr.Interface( | |
fn=Log_reg_predict, | |
inputs=["text", "text", "text", "number", "number"], | |
outputs=[ | |
gr.Label(label="Model Prediction"), | |
gr.Image(label="Shapley Values"), | |
], | |
title="Logistic Regression Profit Prediction", | |
description="Create your own purchases and see if the Logistic Regression model predicts they will make above or below the median profit\n\nValid regions: ['Central', 'East', 'South', 'West']\n\nValid product categories: ['Furniture', 'Office Supplies', 'Technology']\n\nValid product sub-categories: ['Accessories', 'Appliances', 'Art', 'Binders', 'Bookcases', 'Chairs', 'Copiers', 'Envelopes', 'Fasteners', 'Furnishings', 'Labels', 'Machines', 'Paper', 'Phones', 'Storage', 'Supplies', 'Tables']", | |
) | |
SVM_tab = gr.Interface( | |
fn=SVM_predict, | |
inputs=["text", "text", "text", "number", "number"], | |
outputs=[ | |
gr.Label(label="Model Prediction"), | |
gr.Image(label="Shapley Values"), | |
], | |
title="SVM Profit Prediction", | |
description="Create your own purchases and see if the SVM model predicts they will make above or below the median profit\n\nValid regions: ['Central', 'East', 'South', 'West']\n\nValid product categories: ['Furniture', 'Office Supplies', 'Technology']\n\nValid product sub-categories: ['Accessories', 'Appliances', 'Art', 'Binders', 'Bookcases', 'Chairs', 'Copiers', 'Envelopes', 'Fasteners', 'Furnishings', 'Labels', 'Machines', 'Paper', 'Phones', 'Storage', 'Supplies', 'Tables']", | |
) | |
XGB_tab = gr.Interface( | |
fn=XGB_predict, | |
inputs=["text", "text", "text", "number", "number"], | |
outputs=[ | |
gr.Label(label="Model Prediction"), | |
gr.Image(label="Shapley Values"), | |
], | |
title="XGB Profit Prediction", | |
description="Create your own purchases and see if the XGB model predicts they will make above or below the median profit\n\nValid regions: ['Central', 'East', 'South', 'West']\n\nValid product categories: ['Furniture', 'Office Supplies', 'Technology']\n\nValid product sub-categories: ['Accessories', 'Appliances', 'Art', 'Binders', 'Bookcases', 'Chairs', 'Copiers', 'Envelopes', 'Fasteners', 'Furnishings', 'Labels', 'Machines', 'Paper', 'Phones', 'Storage', 'Supplies', 'Tables']", | |
) | |
demo = gr.TabbedInterface([LogReg_tab, SVM_tab, XGB_tab], tab_names=["Logistic Regression", "SVM", "XGB"], theme=gr.themes.Soft()) | |
demo.launch(debug=True) |